AI marketing automation

Two agents.
Zero headcount.

Cold outbound and content marketing — fully automated. LinkedIn to inbox. Transcript to post schedule. You set the ICP and offer. The agents run the rest.

See how it works ← Origin Forge
Agent 01
Cold Outbound
LinkedIn creator monitoring → ICP filter → email enrichment → AI-personalised copy → send via Instantly + HeyReach
Agent 02
Content Engine
Upload any transcript → extract insights → write LinkedIn posts → schedule via Ordinal → snowball winning topics
Cold outbound ◆
LinkedIn scraping ◆
ICP filtering ◆
Email enrichment ◆
Instantly ◆
HeyReach ◆
AI copywriting ◆
Transcript → posts ◆
Ordinal scheduling ◆
Snowball winners ◆
Cold outbound ◆
LinkedIn scraping ◆
ICP filtering ◆
Email enrichment ◆
Instantly ◆
HeyReach ◆
AI copywriting ◆
Transcript → posts ◆
Ordinal scheduling ◆
Snowball winners ◆

Agent 01 — Cold Outbound

From LinkedIn
to inbox.

Monitor any creator's posts. Extract everyone who engages. Score against your ICP. Enrich their email. Write a personalised message. Send automatically.

01
Monitor creator posts
Add LinkedIn creators to your watchlist. The agent fetches their latest posts and extracts every person who liked, commented, or reposted.
Apify scraper
02
ICP scoring
Each engager is scored against your ICP definition — job title, industry, and company size. Non-fits are discarded before any enrichment spend.
icp.json config
03
Waterfall email enrichment
ICP fits go through a multi-source waterfall: Apollo → Hunter → Clearbit. Emails are then validated to keep bounce rates low.
Apollo · Hunter · Clearbit
04
AI-written personalised copy
Claude writes a personalised cold email and LinkedIn DM for each lead — using their role, company, and the post they engaged with as context. Not a template.
Claude AI
05
Send on both channels
Emails go through Instantly campaigns. LinkedIn DMs go via HeyReach. Positive replies are detected and Claude drafts follow-ups automatically.
Instantly · HeyReach

Both agents

Cold pipeline
meets content flywheel.

Each agent runs independently. Together they cover outbound and inbound.

Agent 01

Cold
Outbound

Finds and emails your ideal customers by watching where they already spend time — creator posts in your niche — then reaching out at the moment of highest intent.

→Watch creators whose audience is your ICP
→Extract engagers from their latest posts
→Score against ICP — title, industry, size
→Enrich email via Apollo / Hunter waterfall
→Write copy personalised per lead with Claude
→Send email + DM via Instantly + HeyReach
→Handle replies — Claude drafts follow-ups
Agent 02

Content
Engine

Turns any source material — podcast, sales call, interview, notes — into a weeks-long LinkedIn post schedule. Finds what performs, then writes more of it.

→Upload transcript — any .txt source material
→Extract insights — Claude pulls the key ideas
→Write posts — one post per insight, your tone
→Schedule via Ordinal — 1 post/day at 09:00
→Analyse performance — impressions per post
→Snowball winners — remix top themes into new posts

Command line

One command
per agent.

No dashboard. No SaaS subscription. Configure your ICP and offer once, then run.

# Cold outbound — full pipeline
python run.py outbound --offer "We build AI agents for SaaS marketing teams."

# Dry run — preview leads without sending
python run.py outbound --dry-run --max 25

# Content engine — transcript → scheduled posts
python run.py content --transcript data/podcast_transcript.txt

# Handle positive replies from Instantly inbox
python run.py replies --goal "book a demo"

# Analyse top posts and snowball winners
python run.py snowball --account ord_abc123

# Add a creator to watch
python run.py creators --add https://linkedin.com/in/example